Bayesian Product Discovery

Vanity metrics feel like control. They don’t make decisions easier.

This tool is for disciplined mind-changing: start with an outside-view prior, then update it when a real test comes back.

Each row is one binary test. Pick what happened (positive / negative), then enter the test’s sensitivity (true positive rate) and specificity (true negative rate).

Step 1 — Prior \(P(H)\) (Outside View) 0.2 Baseline success rate from a reference class (your outside view). Or type a value in \([0,1]\).

Hypothesis \(H\) Plain-English hypothesis you’re investing in.

Step 2 — Signals (tests) Add tests that would actually change a decision

Prior

0.20

Posterior (after last test)

—

Number of tests

Steps

Step	Observed	Sensitivity \(P(E^+\mid H)\)	Specificity \(P(E^-\mid \neg H)\)	LR (observed)	\(P(H)\) after update

Step 3 — Posterior (Update)

P(H|E) = (P(H)·P(E|H)) / (P(H)·P(E|H) + (1−P(H))·P(E|¬H))